Date: Mon, 22 Jun 2009 13:55:03 +0200
From: Ingo Molnar <mingo@elte.hu>
To: eranian@gmail.com
Cc: LKML <linux-kernel@vger.kernel.org>,
       Andrew Morton <akpm@linux-foundation.org>,
       Thomas Gleixner <tglx@linutronix.de>,
       Robert Richter <robert.richter@amd.com>,
       Peter Zijlstra <a.p.zijlstra@chello.nl>,
       Paul Mackerras <paulus@samba.org>, Andi Kleen <andi@firstfloor.org>,
       Maynard Johnson <mpjohn@us.ibm.com>, Carl Love <cel@us.ibm.com>,
       Corey J Ashford <cjashfor@us.ibm.com>,
       Philip Mucci <mucci@eecs.utk.edu>, Dan Terpstra <terpstra@eecs.utk.edu>,
       perfmon2-devel <perfmon2-devel@lists.sourceforge.net>
Subject: Re: I.9 - Group reading
Message-ID: <20090622115503.GJ24366@elte.hu>
References: <7c86c4470906161042p7fefdb59y10f8ef4275793f0e@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <7c86c4470906161042p7fefdb59y10f8ef4275793f0e@mail.gmail.com>
User-Agent: Mutt/1.5.18 (2008-05-17)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 933
Lines: 22

> 9/ Group reading
>
> It is possible to start/stop an event group simply via ioctl() on
> the group leader. However, it is not possible to read all the
> counts with a single with a single read() system call. That seems
> odd. Furhermore, I believe you want reads to be as atomic as
> possible.

If you want an atomic snapshot you can do it: disable the group,
read out the counts, enable the group.

But, as your other comment under I/5 indicates, there are ways to
read out the PMU directly, via RDPMC instructions. Those are not
atomic either if used for multiple counters. Is your argument that
they are thus useless?

But if there is a strong use-case we can add PERF_FORMAT_GROUP.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/